C. Carson, S. Belongie, H. Greenspan, and J. Malik. Regionbased image querying. In Proc. IEEE Workshop on ContentBased Access of Image and Video Libraries, pages 42--49, 1997.  Home/Search   Document Details and Download   Summary   Related Articles   Check

....systems. One approach is to define structural features that attempt to capture the structure of a class of images [10] 11] Another is to segment images into regions whose features are typical of certain well known objects, such as tigers and zebras, which have characteristic color and texture [2] or characteristic shape [3] Queries can request images that have regions with certain properties in certain spatial relationships [9] Another approach [8] is to employ user relevance feedback to refine the query results, but this is usually paired with the query by example approach. Our ....

C. Carson, S. Belongie, H. Greenspan, J. Malik, "Regionbased Image Querying," Proceedings of the 1997.

....experimental results for an aerial video dataset are presented in Section 5. Section 6 concludes with a discussion. 2 Related Work Several approaches to applying association rules to image datasets have been proposed. Ordonez and Omiecinksi [2] use segmentation results from the Blobworld system [3] to mine the co occurrence of image regions that have been labeled as similar using an empirically determined distance measure and threshold. The segmented regions are viewed as items and the images are viewed as transactions so that the resulting rules are of the form, The presence of regions A ....

C. Carson, S. Belongie, H. Greenspan, and J. Malik, "Region-based image querying," in Proc. CVPR '97 CBAIVL Workshop, 1997.

....degrees of freedom. As a result there is an overwhelming amount of different techniques that have been applied to image retrieval in recent years: boosting [136] clustering [99] edge extraction [54] grouping [36] hidden Markov models [82] various histograms [33, 133, 107] image segmentation [3, 20, 72, 146], invariant features [111, 64, 41, 27] keypoint extraction [111] moments [29, 74] motion estimation [2] probabilistic matching [1, 140, 139] self organizing maps [67] shape matching [64, 68] texture features [33, 70, 43] transportation problem solving [101] and wavelets [59, 74] just to ....

C. Carson, S. Belongie, H. Greenspan, and J. Malik. Region-based image querying. In IEEE Workshop on Content-based Access of Image and Video Libraries, pages 42--49, San Juan, Puerto Rico, June 1997.

....the good work done by the color features. Several techniques have been reported in the literature to compute the texture in a pixel neighborhood. The famous ones include Multiresolution Simultaneous Autoregressive (MSAR) model [13] Gabor Wavelets [14] and the Second Moment Eigenstructure (SME) [15, 16]. In the present work, we have used a weaker measure of texture yielded by the Second Moment Eigenstructure (SME) which can capture the essential neighborhood characteristics of a pixel. The second moment matrix at each image pixel (i, j) is given by: # # # # # # # # # # ## ## cW cW ....

....moment matrix can be shown to be a modification of bilinear, symmetric positive definite metric defined over the 2D image manifold embedded in a 5D space of (R, G, B, i, j) 16] The eigenstructure of the second moment matrix represents the textural properties. Two measures have been defined in [15] using the eigenvalues of the matrix, a) anisotropy = 1 k I 2 1 , and (b) normalized strength = 2 ( 1 2 ) where 1 and 2 are the two eigenvalues of matrix M(i, j) and 1 2 . In the present work, we have used the combination of anisotropy and the strength called texture ....

C. Carson, S. Belongie, H. Grennspan, and J. Malik, "Region-Based Image Querying", CVPR'97, Workshop on Content-Based Access of Image and Video Libraries, 1997.

....segmentation results in a conspicuous region with isolated points as its limit case [13, 7] Weak segmentation is a necessity in achieving occlusion or clutter invariant descriptions. From the weakly segmented regions, a selection is made on their salience. The most salient regions are stored [2]. As the information of the image is condensed into just a limited number of feature values, the information should be selected with precision and proven robustness. In [7, 5] photometric invariance is the leading principle in summarizing the image in salient transitions in the image. 6. ....

....of some feature set. The similarity between histograms has been measured by the intersection distance in [17] a necessity for invariance to occlusion. When following the line of salient features, invariant similarity can be performed by uniquely identifying the feature signature of the blob [2] or one can store all salient points from one image in a histogram on the basis of a few characteristics, such as invariant color on the inside and outside. The similarity is then based on the group wise presence of enough similar points. To that end, vector space modeling of NLP implies the ....

C. Carson, S. Belongie, H. Greenspan, and J. Malik. Region-based image querying. In Proceedings of the IEEE International Workshop on Content-Based Access of Image and Video Databases, 1997.

.... are: 1) The spatial structures (e.g. 5, 8, 10, 15, 16, 17, 18] Different structural elements (e.g. buildings, road, tables, walls, with particular orientation patterns, smoothness roughness) compose each context (e.g. rooms, streets, shopping center) 2) The spatial organization (e.g. [2, 8, 16, 17]) The structural elements have particular spatial arrangements. Each context imposes certain organization laws. 3) The color distribution [2, 5, 8, 15, 18] As described below, we propose a low dimensional holistic representation that encodes the structural scene properties [10] Color is not ....

....orientation patterns, smoothness roughness) compose each context (e.g. rooms, streets, shopping center) 2) The spatial organization (e.g. 2, 8, 16, 17] The structural elements have particular spatial arrangements. Each context imposes certain organization laws. 3) The color distribution [2, 5, 8, 15, 18]. As described below, we propose a low dimensional holistic representation that encodes the structural scene properties [10] Color is not taken into account in this study, although the framework can be extended to include this attribute. Figure 2. The first three PCs of the WFT at 16x16 ....

Carson, C., Belongie, S., Greenspan, H., and Malik, J. 1997. Region-based image querying. Proc. IEEE W. on ContentBased Access of Image and Video Libraries, pp: 42--49.

.... from the USGS GNIS data set (D = 2) USGS95] This is a national version of the Sequoia 2000 storage benchmark [STON93] Spatial coordinates plus time from NOAA s GTSPP data set (D = 4) HAMI94] Image feature vectors from the Berkeley Digital Library Project s Blobworld system (D = 20) [CARS97]. The 20 dimensions result from applying the singular value decomposition to 256 bin histogram values in the CIE LUV Dimensionality Density Insertion Bulk load random Hilbert Hilbert STR DD 0 D 2 Data set Records GNIS 1.837 1.746 3.997 2.773 2.274 1.627 Uni2 2 2 31.33 4.894 2.974 2.553 ....

C. Carson, S. Belongie, H. Greenspan and J. Malik, "Region-Based Image Querying," Proc. IEEE Wksp. on ContentBased Access of Image and Video Libraries, San Juan, Puerto Rico, June 1997, 42-49.

....10 results in Table 2 that show 50 to more than 100 increase in the position of the 10th category image, between theGMM and histogram measures. Among the histogram distance measures, the Euclidean is very clearly the worst (as expected) with the Disc. KL measure quite consistently the best. 5. 3 The combined GMM KL framework in category modeling and category matching The final set of experiments deals with the concept of category modeling and matching. An example of category modeling was shown in Figure 4. It is of interest to investigate the following questions: are the category models ....

....of category modeling was shown in Figure 4. It is of interest to investigate the following questions: are the category models representative of the underlying image set Can image matching enable image classification An initial investigation was conducted with the results listed in Tables 3 and 4 and plotted in Figure 10. Table 3 lists distances between category models, following equation (15) In the field category, for example, the closest category model is seen to be the waterfall category (see Figure 5) Similar relationships may be learned automatically by the system, and ....

[Article contains additional citation context not shown here]

C. Carson, S. Belongie, H. Greenspan, and J. Malik. Region-based image querying. In Proc. of the IEEE Workshop on Content-based Access of Image and Videolibraries (CVPR'97), pages 42--49, 1997

....evaluated and compared in [1, 2, 3] Several works extend the histogram representation to include spatial information. Unsupervised segmentation of an image into homogeneous regions in the feature space, such as the color and texture space, can be found in the blobworld image representation [4, 5]. In [5] the user composes a query by viewing the blobworld representation, and selecting the blobs to match, along with possible weighting of the blob features. In essence, the image matching problem is shifted to a (one or two) blob to image matching problem. Each blob in one image is compared ....

....region EMD matching system. rameters of a mixture of k Gaussians in the feature space. The first step in applying the EM algorithm to the problem at hand is to initialize the mixture model parameters. The K means algorithm is utilized to extract the data driven initialization. The MDL principle [4] is used to select the number of mixture components (or number of means) k, as best suits the natural number of groups present in the image. The EM algorithm, along with the model selection, can be applied to a particular feature space, such as color (in which case the image color space is being ....

C. Carson, S. Belongie, H. Greenspan and J. Malik, "Regionbased Image Querying ", Proc. of the IEEE Workshop on Content-based Access of Image and Video libraries (CVPR'97) pp 42-49, 1997.

No context found.

C. Carson, S. Belongie, H. Greenspan, and J. Malik. Regionbased image querying. In Proc. IEEE Workshop on ContentBased Access of Image and Video Libraries, pages 42--49, 1997.

No context found.

C. Carson, S. Belongie, H. Greenspan, J. Malik, "Regionbased Image Querying," Proceedings of the 1997.

No context found.

CARSON C., BELONGIE S., GREENSPAN H., MALIK J., "Region-Based Image Querying", Proceedings of the IEEE Worshop on Content-Based Access of Image and Video Libraries (CVPR'97), 1997.

No context found.

C. Carson, S. Belongie, H. Greenspan, and J. Malik, "Region-based image querying," in Proceedings of the IEEE Workshop on Content-Based Access of Image and Video Li- braries (IEEE Computer Society Press, Los Alamitos, Calif., 1997), pp. 42--49.

No context found.

Carson C., Belongie S., Greenspan H. and Malik J. (1997). Region-based Image Querying. In CVPR'97 Workshop on Content-based Access to Image and Video Libraries.

No context found.

C. Carson, S. Belongie, H. Greenspan, and J. Malik, "Region-based image querying," in Proceedings of the IEEE International Workshop on Content-Based Access of Image and Video Databases (1997).

No context found.

C. Carson, S. Belongie, H. Greenspan, J. Malik, Region--based image querying, in: CPR Workshop on Content--Based Access of Image and Video Libraries, 1997.

No context found.

C. Carson, S. Belongie, H. Greenspan, J. Malik, Region--based image querying, in: Proceedings of the 1997.

No context found.

C. Carson, S. Belongie, H. Greenspan, and J. Malik. Region-based image querying. In Proc. IEEE Workshop on Content-based Access of Image and Video Libraries, 1997.

No context found.

Carson C, Belongie S, Greenspan H & Malik J (1997) Region-based image querying. Proc. IEEE Workshop on Content-Based Access of Image and Video Libraries, San Juan, Puerto Rico, 42-49.

No context found.

C. Carson, S. Belongie, H. Greenspan, and J. Malik, Region based image querying," IEEE Proc. CBAIVL'97, pp. 42-49, 1997.

No context found.

C. Carson et al., Region-Based Image Querying, In Proceedings of CVPR'97 Workshop on Content-based Access to Image and Video libraries (CAIVL'97), (1997).

No context found.

C. Carson, S. Belongie, H. Greenspan, and J. Malik. Region-based image querying. In Proc. CVPR Workshop on Content-Based Access of Image and Video Libraries, 1997.

No context found.

Chad Carson, Serge Belongie, Hayit Greenspan, and Jitendra Malik. Region-based image querying, In Proc. IEEE Workshop on Content-based access of Image and Video Libraries, in conjunction with IEEE CVPR, 1997.

No context found.

C. Carson, S. Belongie, H. Greenspan, and J. Malik, "Region-based image querying," in CVPR'97 Workshop on Content-Based Access of Image and Video Libraries, 1997.

No context found.

Carson, C., Belongie, S., Greenspan, H., and Malik, J. 1997. Region-based image querying. Content-Based Access of Image and Video Libraries, pp: 42 49. Proc. IEEE W. on

First 50 documents  Next 50

Online articles have much greater impact   More about CiteSeer.IST   Add search form to your site   Submit documents   Feedback  

CiteSeer.IST - Copyright Penn State and NEC